Detection Event Log¶
Class: DetectionEventLogBlockV1
Source: inference.core.workflows.core_steps.analytics.detection_event_log.v1.DetectionEventLogBlockV1
This block maintains a log of detection events from tracked objects. For each tracked object it records: class name, first and last seen frame numbers, absolute wall-clock timestamps (Unix epoch floats derived from frame_timestamp metadata, or time.time() as fallback), and relative timestamps in seconds since the video started. Objects must be seen for a minimum number of frames (frame_threshold) before being moved from 'pending' to 'logged' status. Stale events (not seen for stale_frames frames) are removed during periodic cleanup (every flush_interval frames). When a logged event goes stale it is emitted in the complete_events output, which contains the full event data for objects that were tracked long enough to be logged and have since left the scene. The reference_timestamp parameter is deprecated and no longer used.
Type identifier¶
Use the following identifier in step "type" field: roboflow_core/detection_event_log@v1to add the block as
as step in your workflow.
Properties¶
| Name | Type | Description | Refs |
|---|---|---|---|
name |
str |
Enter a unique identifier for this step.. | ❌ |
frame_threshold |
int |
Number of frames an object must be seen before being logged.. | ✅ |
flush_interval |
int |
How often (in frames) to run the cleanup operation for stale events.. | ✅ |
stale_frames |
int |
Remove events that haven't been seen for this many frames.. | ✅ |
reference_timestamp |
float |
Deprecated, no longer used. Absolute timestamps are now taken directly from frame_timestamp metadata (or time.time() as fallback).. | ✅ |
fallback_fps |
float |
Fallback FPS to use when video metadata does not provide FPS information. Used to calculate relative timestamps.. | ✅ |
The Refs column marks possibility to parametrise the property with dynamic values available
in workflow runtime. See Bindings for more info.
Runtime compatibility¶
-
soft— runtimehosted_serverless,dedicated_deployment; executionremote; inputvideo - Block keeps per-video state in process memory (keyed by video_metadata.video_identifier). With remote step execution on stateless or multi-replica HTTP runtimes, successive requests may be served by different worker processes, so the state resets between calls and the output is meaningless for tracking / counting / aggregation. Use local step execution in an InferencePipeline for stable cross-frame results.
-
soft— inputimage - Block depends on temporal context from video or repeated-frame workflows. With a still image/photo, there is no meaningful history to track, compare, aggregate, or visualize, so the block provides little or no benefit.
Available Connections¶
Compatible Blocks
Check what blocks you can connect to Detection Event Log in version v1.
- inputs:
Detections Filter,Line Counter,Time in Zone,Path Deviation,Object Detection Model,Instance Segmentation Model,Detection Offset,ByteTrack Tracker,Detection Event Log,Per-Class Confidence Filter,YOLO-World Model,SAM 3 Interactive,Detections Transformation,Detections Consensus,Byte Tracker,Perspective Correction,SAM 3,Mask Area Measurement,Google Vision OCR,Detections Merge,Dynamic Zone,Detections Combine,Instance Segmentation Model,Camera Focus,Detections List Roll-Up,Mask Edge Snap,PTZ Tracking (ONVIF),Moondream2,Object Detection Model,VLM As Detector,EasyOCR,Bounding Rectangle,SAM2 Video Tracker,Overlap Filter,SAM 3,Time in Zone,Detections Classes Replacement,Instance Segmentation Model,Byte Tracker,Gaze Detection,Detections Stabilizer,SORT Tracker,Instance Segmentation Model,Velocity,Track Class Lock,Byte Tracker,Object Detection Model,Time in Zone,Cosine Similarity,Identify Changes,Dynamic Crop,Detections Stitch,Path Deviation,Template Matching,BoT-SORT Tracker,SAM3 Video Tracker,Motion Detection,Camera Focus,OC-SORT Tracker,SAM 3,Segment Anything 2 Model,VLM As Detector,Seg Preview,OCR Model - outputs:
Line Counter,Trace Visualization,Path Deviation,Image Stack,Anthropic Claude,Per-Class Confidence Filter,Icon Visualization,SIFT Comparison,Morphological Transformation,Color Visualization,Perspective Correction,Corner Visualization,Roboflow Custom Metadata,Detections Merge,Halo Visualization,Dynamic Zone,Keypoint Detection Model,Email Notification,Halo Visualization,Object Detection Model,Background Color Visualization,Ellipse Visualization,Email Notification,Twilio SMS/MMS Notification,Text Display,Polygon Visualization,Crop Visualization,Absolute Static Crop,Image Preprocessing,Model Monitoring Inference Aggregator,Florence-2 Model,Motion Detection,Heatmap Visualization,Detections Filter,Blur Visualization,Instance Segmentation Model,Stability AI Outpainting,Anthropic Claude,PLC EthernetIP,Background Subtraction,Keypoint Visualization,Webhook Sink,Byte Tracker,Stitch Images,Florence-2 Model,Detections List Roll-Up,Mask Edge Snap,Line Counter,Triangle Visualization,Slack Notification,Overlap Filter,Time in Zone,Detections Stabilizer,Image Contours,Keypoint Detection Model,Pixel Color Count,Roboflow Asset Library Attributes,Image Slicer,Time in Zone,Stitch OCR Detections,Line Counter Visualization,Image Threshold,Distance Measurement,QR Code Generator,Detection Offset,ByteTrack Tracker,Detection Event Log,Detections Transformation,Microsoft SQL Server Sink,Mask Area Measurement,Twilio SMS Notification,Image Blur,Detections Combine,Morphological Transformation,Size Measurement,Roboflow Vision Events,PTZ Tracking (ONVIF),Stability AI Inpainting,Classification Label Visualization,Stitch OCR Detections,SAM2 Video Tracker,Bounding Rectangle,Event Writer,Grid Visualization,Byte Tracker,Mask Visualization,Dominant Color,Image Slicer,Velocity,Label Visualization,Identify Outliers,Reference Path Visualization,Byte Tracker,SIFT Comparison,OPC UA Writer Sink,Dot Visualization,Identify Changes,Path Deviation,Detections Stitch,Dynamic Crop,Circle Visualization,BoT-SORT Tracker,Camera Focus,Segment Anything 2 Model,Overlap Analysis,Object Detection Model,SAM 3 Interactive,Detections Consensus,Bounding Box Visualization,Instance Segmentation Model,Roboflow Visual Search,Roboflow Dataset Upload,SAM 3,Detections Classes Replacement,Instance Segmentation Model,Pixelate Visualization,Keypoint Detection Model,Roboflow Dataset Upload,SORT Tracker,Instance Segmentation Model,Track Class Lock,Anthropic Claude,Object Detection Model,Time in Zone,MQTT Writer,Polygon Visualization,OC-SORT Tracker,Model Comparison Visualization
Input and Output Bindings¶
The available connections depend on its binding kinds. Check what binding kinds
Detection Event Log in version v1 has.
Bindings
-
input
image(image): Reference to the image for video metadata (frame number, timestamp)..detections(Union[object_detection_prediction,instance_segmentation_prediction]): Tracked detections from byte tracker (must have tracker_id)..frame_threshold(integer): Number of frames an object must be seen before being logged..flush_interval(integer): How often (in frames) to run the cleanup operation for stale events..stale_frames(integer): Remove events that haven't been seen for this many frames..reference_timestamp(float): Deprecated, no longer used. Absolute timestamps are now taken directly from frame_timestamp metadata (or time.time() as fallback)..fallback_fps(float): Fallback FPS to use when video metadata does not provide FPS information. Used to calculate relative timestamps..
-
output
event_log(dictionary): Dictionary.detections(Union[object_detection_prediction,instance_segmentation_prediction]): Prediction with detected bounding boxes in form of sv.Detections(...) object ifobject_detection_predictionor Prediction with detected bounding boxes and segmentation masks in form of sv.Detections(...) object ifinstance_segmentation_prediction.total_logged(integer): Integer value.total_pending(integer): Integer value.complete_events(dictionary): Dictionary.
Example JSON definition of step Detection Event Log in version v1
{
"name": "<your_step_name_here>",
"type": "roboflow_core/detection_event_log@v1",
"image": "$inputs.image",
"detections": "$steps.byte_tracker.tracked_detections",
"frame_threshold": 5,
"flush_interval": 30,
"stale_frames": 150,
"reference_timestamp": 1726570875.0,
"fallback_fps": 1.0
}